How many students should I survey?

This tool can be used for working out how many people you need to survey to gain statistically useful results. http://www.raosoft.com/samplesize.html

To simplify things we’ve put together a rough guide for you with some typical values of confidence and tolerance:

If you have this many students	You should aim to survey this many
100	80
500	218
1,000	278
5,000	357
10,000	370
15,000	375
30,000	380

To illustrate:

 

How can I tell if two numbers are significantly different?

Significance is a way of saying that a difference between numbers is not simply due to chance.

For illustration, if you toss 5 coins you can expect them all to land on the same side every roughly 6 times out of 100. Significance tells you how many times you would need to see that effect before you can start to shout about how you have magic coins (or indeed weighted ones).

You can use this tool to check it:

http://www.harrisresearchpartners.com/SigDiffCalculator.htm

For instance in a survey with 370 respondents, where you are comparing the response rate of 170 people with the response rate of the other 200, you would need to show a difference of 10.13% to claim significance.

 

How do I work with POLAR data?

Participation of Local Areas (POLAR) data can be useful as a rough proxy for social class. It ranks a postcode from 1-5 based on the number of people who go to university.

For more detail and to download the list of postcodes and POLAR indices visit the HEFCE website:

http://www.hefce.ac.uk/analysis/yp/POLAR/Map,of,young,participation,areas/

The dataset is unfortunately too large to open in Microsoft Excel, plus it would be ridiculously time-consuming to go through your student database one-by-one working out what POLAR index each postcode has. This method will take your list of postcodes and pull out POLAR index for each one:

http://www.nusconnect.org.uk/resources/how-to-use-the-polar-lookup